What the work actually involves
This is applied-software work, not writing work. You will be asked to carry out realistic Word tasks on your own Mac and capture them in a way an AI model can learn from: screenshots at full resolution, step-by-step records of what you clicked and why, and notes on where the application behaves unexpectedly. Typical assignments mirror the work of legal, policy, consulting and technical-writing professionals — rebuilding a broken style hierarchy, repairing cross-references and a table of contents after a heavy edit, reconciling multi-author tracked changes, setting up section breaks with differing headers, managing citations, or preparing a document for export and comparison.
A second strand of the work is evaluation: reviewing an AI agent's attempt at one of these tasks and judging whether it reached the correct end state, whether it took a route a competent professional would take, and where exactly it went wrong. macOS matters here. Word for Mac diverges from Windows in ribbon layout, keyboard shortcuts, styles pane behaviour and several referencing features, and the project needs people who know those differences from use rather than from documentation.
What the platform screens for
Mercor's screen is AI-led and follow-up heavy. Expect it to probe for named menus, panes and dialogs rather than general claims of proficiency, and to ask what you do when Word misbehaves — a corrupted numbering list, a TOC that will not update, styles imported from a pasted section. It also verifies the hard logistics: Word installed and openable today on macOS, and a display meeting the capture threshold. Claims that fail a screenshot check end the process.
Logistics
- Fully remote and asynchronous; you pick up tasks from a queue rather than working fixed shifts.
- No software, licences, VMs or hardware are provided — your own Mac and your own Word install only.
- Display must clear ~2.5 MP, either as one high-resolution monitor (QHD and up, or a 14"/16" MacBook Pro built-in panel) or as two 1920×1080+ monitors captured as a single spanning image.
- Volume varies by project phase; contributors commonly report a handful of hours a week rather than full-time load. Pay of $60–70/hr reflects observed rates on this listing, not a guarantee.